The amount of soil ingested by young children was estimated by measuring the titanium, aluminum, and acid-insoluble residue in soil and feces. As intake of each of these tracers is also possible from sources other than soil ingestion, the amount of soil ingested was estimated to be not higher than the lowest of the three separate estimates. This estimate, the limiting tracer method (LTM) value, was then corrected for the similarly calculated mean LTM value for a group of hospitalized children without access to soil and dust. The study groups included children in three different environmental situations: day-care centers, campgrounds, and hospitals. The day-care center groups were sampled twice. From these groups, 162 children produced usable feces samples during both sampling periods. The camping groups and the hospitalized (control) group were sampled once. For the day-care center groups, the estimated geometric mean soil intake varied from 0 to 90 mg/day and for the camping groups these estimates ranged from 30 to 200 mg/day (in dry weight). Using estimates of the "true" between-child GSD values, the 90th percentile of the estimated soil intakes was shown to be typically 40-100 mg/day higher than the geometric means of these estimates. In the day-care center groups few correlations with the geometric mean LTM values were found for variables concerning living conditions, mouthing behavior, playing habits, etc. A strong correlation was found with weather. During dry weather the younger children especially showed higher LTM values. In the camping group the weather also influenced the mean LTM value only in the younger age groups. Analysis of variance showed that a single LTM value of a child has a low predictive value with regard to the LTM value of the next few days or that of a few months later. Therefore it seems reasonable to use group statistics as estimates of soil ingestion in health risk assessments of soil pollution incidents.